Yan Gai and Laurel H. Carney

Transcription

1 Yan Gai and Laurel H. Carney J Neurophysiol 99: , First published Jan 16, 2008; doi: /jn You might find this additional information useful... This article cites 50 articles, 19 of which you can access free at: Updated information and services including high-resolution figures, can be found at: Additional material and information about Journal of Neurophysiology can be found at: This information is current as of March 12, Journal of Neurophysiology publishes original articles on the function of the nervous system. It is published 12 times a year (monthly) by the American Physiological Society, 9650 Rockville Pike, Bethesda MD Copyright 2005 by the American Physiological Society. ISSN: , ESSN: Visit our website at

2 J Neurophysiol 99: , First published January 16, 2008; doi: /jn Influence of Inhibitory Inputs on Rate and Timing of Responses in the Anteroventral Cochlear Nucleus Yan Gai 1,2 and Laurel H. Carney 1,2,3 1 Department of Biomedical and Chemical Engineering, 2 Institute for Sensory Research, and 3 Department of Electrical Engineering and Computer Science, Syracuse University, Syracuse, New York Submitted 26 January 2007; accepted in final form 15 January 2008 Gai Y, Carney LH. Influence of inhibitory inputs on rate and timing of responses in the anteroventral cochlear nucleus. J Neurophysiol 99: , First published January 16, 2008; doi: /jn Anatomical and physiological studies have shown that anteroventral cochlear nucleus (AVCN) neurons receive glycinergic and GABAergic inhibitory inputs. In this study, changes in the temporal responses of AVCN neurons to pure tones and complex sounds after blocking inhibition were analyzed. Blocking inhibition influenced the temporal responses of each type of AVCN neuron. Choppers showed more chopping peaks and shortened chopping cycles after blocking inhibition. Sustained and slowly adapting choppers showed increased regularity throughout the response duration after blocking inhibition, whereas most transient choppers showed increased regularity in the early part of the response. Diverse changes in temporal response patterns were observed in neurons with primary-like and unusual responses, with several neurons showing a large decrease in the first-spike latency after blocking inhibition. This result disagreed with previous findings that onset responses are less affected than sustained responses by manipulating inhibition. Although blocking inhibition had a greater effect on spontaneous activity than that on tone-evoked activity, the change in spontaneous activity was less significant because of larger variability. In addition, for relatively high level masker noises, blocking inhibition had similar effects on responses to noise-alone and noise-plus-tone stimuli, in contrast with previous studies with low-level background noise. In general, inhibition had an enhancing effect on temporal contrast only for responses to amplitude-modulated tones, for which envelope synchrony was enhanced. Results of this study contribute new information about the characteristics, functional roles, and possible sources of inhibitory inputs received by AVCN neurons. INTRODUCTION The cochlear nucleus (CN) sends major inputs to higher levels of the auditory pathway. Understanding the response properties and information-processing mechanisms of CN neurons can aid the study of higher-level neurons. Established models of auditory-nerve fibers (ANFs; Giguère and Woodland 1994; Robert and Eriksson 1999; Zhang et al. 2001; Zilany and Bruce 2006) provide useful tools to simulate the responses of CN neurons caused by excitatory ANF inputs. However, it is difficult to understand the role of inhibitory inputs because the source of inhibition remains unclear for specific neurons or neuron types. All principal cell types in the anteroventral cochlear nucleus (AVCN) receive inputs that stain for glycine (Wenthold et al. 1988) and -aminobutyric acid (GABA; Saint Marie et al. Address for reprint requests and other correspondence: L. H. Carney, Departments of Biomedical Engineering and Neurobiology and Anatomy, University of Rochester, Box 603, 601 Elmwood Ave., Rochester, NY ( laurel.carney@rochester.edu). 1989). Possible glycinergic inputs include projections from the tuberculoventral (vertical) neurons in the dorsal cochlear nucleus (DCN) and D-stellate neurons in the posteroventral cochlear nucleus (PVCN) and AVCN (Arnott et al. 2004; Oertel and Wickesberg 1993; Smith and Rhode 1989; Wickesberg and Oertel 1988, 1990). Vertical cells respond actively to characteristic frequency (CF) tones and weakly to broadband noise (Gibson et al. 1985), whereas D-stellate cells respond weakly to CF tones but actively to broadband noise (Rhode and Greenberg 1994). A third source of glycinergic inhibition is the descending input from the superior olivary complex (SOC). A study in guinea pig using retrograde labeling combined with immunocytochemistry reports glycinergic projections from the lateral (LNTB) and ventral (VNTB) nuclei of the trapezoid body and the dorsal periolivary nucleus to the CN. Most of these projections are ipsilateral (Ostapoff et al. 1997). There are reciprocal connections between the CN and the posteroventral periolivary nucleus (PVPO): the only known input to the PVPO is the ascending projection from the CN and the only known output of the PVPO is the descending glycinergic projection to the same area of the CN (guinea pig: Helfert et al. 1989; Thompson and Schofield 2000). There are also commissural glycinergic projections from the contralateral CN to the ipsilateral CN (Babalian et al. 2002; Wenthold 1987). Sources of GABAergic inhibition to the AVCN are less clear. The major source of GABAergic inhibition is presumably the descending projection from the SOC. Most of these GABAergic neurons are bilaterally located in the VNTB (Ostapoff et al. 1997). Although there are GABAergic inhibitory interneurons located in the DCN, no connections between these neurons and the VCN have been found. Golgi cells in the superficial granule cell domain are GABAergic (Kolston et al. 1992). It is likely that some AVCN neurons have distal dendrites that receive inputs from these cells (Ferragamo et al. 1998b), since these cells do not project to the VCN itself, but to regions overlying the VCN. However, inhibition is not expected to have a strong effect when the inputs synapse on the distal dendrites of the target neurons. This assumption is confirmed by the fact that GABAergic inhibitory postsynaptic potentials (IPSPs) are insignificant during brain-slice recording (Ferragamo et al. 1998a; Oertel 1983), when inputs from outside the CN are absent. Early studies that used off-cf tones to detect inhibitory sidebands (Goldberg and Brownell 1973; Martin and Dickson 1983; Rhode and Greenberg 1994) might have yielded mis- The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact /08 $8.00 Copyright 2008 The American Physiological Society 1077

3 1078 Y. GAI AND L. H. CARNEY leading results because excitatory and inhibitory areas overlapped in the frequency responses of AVCN neurons. A more direct approach to study the response properties of inhibitory inputs is to make intracellular recordings of IPSPs. The in vivo intracellular study by Paolini et al. (2005) suggested that the difference in discharge regularity between transient and sustained choppers might originate from the relative amount of overlap between excitatory and inhibitory response areas. More specifically, transient choppers have inhibitory response areas that closely match excitatory areas, and this match results in declined firing rate and regularity compared with sustained choppers. Sustained choppers have more lateral inhibition that is less effective in altering the firing rate and regularity in response to CF tones (Paolini et al. 2005). However, the dominance of excitatory postsynaptic potentials (EPSPs) in response to sound stimuli makes it difficult to detect the presence of IPSPs at any time other than after the tone offset. Moreover, if IPSPs were caused by shunting inhibition, they would be hard to detect because the reversal potential of shunting inhibition is close to the membrane resting potential. Iontophoretically applying inhibitory receptor agonists or antagonists provides a better test of the effect of inhibition. Caspary et al. (1979, 1994) injected glycine/gaba-receptor agonists/antagonists in the AVCN and DCN and found that inhibition alters the maximum response or near-cf response for the majority of AVCN neurons, which is consistent with findings in the PVCN (Palombi and Caspary 1992) and the inferior colliculus (IC) (Palombi and Caspary 1996), but inconsistent with the hypothesis of lateral inhibition. Unfortunately, because sustained choppers were not differentiated from transient choppers in Caspary et al. (1994), it is not clear whether this finding agrees or disagrees with that of Paolini et al. (2005), who reported that sustained choppers receive lateral inhibition. In addition to changes in average rate and frequency response, previous studies also focused on changes in the temporal contrast caused by inhibitory inputs. Several features have been observed related to the temporal contrast. First, spontaneous activity is reported to be more suppressed by inhibitory inputs than is sound-evoked activity (Caspary et al. 1979; Ebert and Ostwald 1995a,b). Second, sustained activity is reported to be more suppressed by inhibitory inputs than are onset responses (Ebert and Ostwald 1995a,b; Kopp-Scheinpflug et al. 2002). Third, activity evoked by low-level background noise is reported to be more suppressed by inhibitory inputs than is tone-evoked activity (Ebert and Ostwald 1995a). Fourth, inhibition enhances synchronization to the stimulus envelope in the PVCN and DCN, especially at low- and mid-modulation frequencies (Backoff et al. 1999). All these findings suggest that inhibition enhances the temporal contrast of CN responses to simple and complex sounds. The present study explored the effect of blocking glycinergic and GABAergic 1 inhibition on AVCN responses to short tones, tones in noise, and amplitude-modulated tones. Numerous 1 The GABAergic inhibition described here refers only to GABA A -receptormediated inhibition, not GABA B, although both are present in the VCN (Altschuler et al. 1986). Ebert and Ostwald (1995a) reported that injecting GABA A -receptor antagonist muscimol had significant effects on neurons that responded to GABA injection, whereas injecting the GABA B -receptor antagonist baclofen on the same neurons showed small effects that were not consistent with the GABA effect. repetitions of short tones were used in a detailed analysis of the changes in temporal responses. Relatively high level noise was used to examine the temporal contrast for stimuli at levels similar to those used in psychophysical studies of detection in noise (Costalupes 1983; Zheng et al. 2002). Moreover, the present study injected inhibitory receptor antagonists, whereas some of the observations described earlier were based on the injection of inhibitory receptor agonists. As pointed out by Ebert and Ostwald (1995b), the steady injection of agonists creates a source of tonic inhibition, which can be qualitatively different from sound-evoked inhibition received by the target neuron. In addition to revealing information about the influence of inhibition on temporal contrast, changes in the responses to tones in noise when blocking inhibition can also help identify the source of the inhibitory inputs. As mentioned earlier, vertical cells and D-stellate cells respond differently to noise and to tones. Therefore the different effects of inhibition on responses to tones and noise might provide clues about the types of neurons that provide inhibition to different AVCN response types. METHODS Animal preparation Details of the surgical procedure are described in Gai and Carney (2006). Briefly, Mongolian gerbils (Meriones unguiculatus), 2 to 4 mo of age, were anesthetized with ketamine and xylazine and placed in a double-walled soundproof booth. A patch of skin was removed from the top of the animal s head and the skull was glued to a head bar mounted in a stereotaxic instrument (model 900, Kopf Instruments) using dental acrylic to ensure long-duration single-unit recordings. A plastic sound-delivery tube was coupled to the left meatus after the pinna was removed. A hole in the dorsal and caudal bulla was made to expose the temporal bone, in which a second small hole was made for electrode penetration through the dura to the AVCN. The surgical procedures were approved by the Syracuse University Institutional Animal Care and Use Committee. Sound stimuli Sound stimuli were created digitally with Matlab and converted to analog signals with a programmable Tucker Davis Technologies System III. Linear compensation of levels for frequencies from 70 Hz to 10 khz was performed for each sound stimulus, based on a calibration table generated at the beginning of each experiment with an ER-7C probe microphone (Etymōtic Research). Neurons with CFs outside the range of 0.3 to 9 khz were not studied. The sound stimuli were short tones, 100% sinusoidally amplitude modulated (SAM) tones, and tones in broadband noise. The CF, threshold, and spontaneous rate (SR) were measured using an automated tuning curve (Liberman 1978). Rate-level functions of short tones at CF (25-ms duration, 5-ms cosine-squared ramps, repeated every 100 ms at several levels with a 10- or 15-dB step size) were most frequently used to monitor the iontophoretic effect on neural responses. One hundred repetitions were obtained at each tone level for the categorization of response types according to Blackburn and Sachs (1989) and for the study of details of the temporal responses to tones. The frequency response area was studied by varying the frequency and level of 25-ms tones (10 repetitions, 15- or 20-dB step size). At each tone level, the tone frequency swept from below CF to above CF (e.g., 2 octaves below and 1 octave above) in half-octave steps, except when limited by the frequency range of the calibration. Three or four tone levels were studied for each neuron, with one level

4 INHIBITORY EFFECTS ON AUDITORY RESPONSES 1079 slightly above threshold, one level at 70 or 80 db SPL, and one or two intermediate levels. The carrier frequency of the SAM tone was set at the neuron s CF. The level of the carrier was about 20 db above the neuron s pure-tone threshold. The SAM tone had a duration of 600 ms and had 5-ms cosine-squared ramps, and was repeated every second for six to eight trials at each of several modulation frequencies, which were arranged from low to high (e.g., from 16 to 1,024 Hz) in 1-octave steps. The tone-in-noise stimulus was either a broadband Gaussian noise ( khz) or the noise plus a CF tone at various tone levels (20 80 db SPL, 10- or 15-dB step size). The noise spectrum level was 30 db SPL (overall level of 70 db SPL). The tone-in-noise stimulus had a duration of 250 ms and had 10-ms cosine-squared ramps and was repeated every 475 s for 20 to 50 repetitions. Tone levels were randomly arranged in each repetition. The noise was randomly created and frozen for each data set and thus varied across tests (and before and after blocking inhibition). After each tone-in-noise trial, the same tone was repeated without the noise. Recording and iontophoresis Neural recordings were amplified by an AC preamplifier (Grass P55) through a high-impedance input module (Grass HZP). A voltagecrossing criterion was used to discriminate spikes from background noise, and the times of the peaks of spikes were marked as spike times with a resolution of 1 s. Only bipolar action potentials were studied, to limit recordings to AVCN cells rather than fibers of passage (Rhode 1998). Single-neuron extracellular recordings and iontophoretic injections were made with six-barrel piggyback electrodes (Havey and Caspary 1980). The recording electrode (15 50 M ) was attached with epoxy to the injecting barrels, protruding m to their tip. The injecting five-barrel electrode (1.2 mm OD 0.68 mm ID, 10 cm in length; World Precision Instruments) was pulled to a tip; the tip was then manually broken to achieve a diameter of m. Each injecting barrel contained one of the following chemicals: strychnine- HCl (10 20 mm, ph 3 3.5), a glycine-receptor antagonist; bicuculline methiodide (10 mm, ph 3 3.5), a -aminobutyric acid type A (GABA A ) receptor antagonist; gabazine (3 5 mm, ph 3 3.5), a GABA A -receptor antagonist; glycine (500 mm, ph 3.5 4); or GABA (500 mm, ph 4 4.5). Injecting currents ( na per barrel; one or two barrels with the same chemical were used at a time) were generated and monitored by a four-channel current generator (NeuroPhore BH-2, Harvard Apparatus); a negative holding current ( 15 na) was maintained for each barrel when no injections were being made. To avoid the buildup of net charge due to injected current, a balancing barrel (the center barrel of the five-barrel electrode) continuously injected a current that was equal to the sum of the currents in the other barrels with inverse polarity. The balancing barrel and the attached recording barrel were filled with 1 M sodium acetate (Kopp-Scheinflug et al. 2002). Sodium was used to be consistent with extracellular ionic concentrations; acetate was chosen to reduce diffusion and transport number during application of hyperpolarizing current (Müller 1992). Occasionally, one of the injecting barrels was filled with 1 M sodium acetate as a control barrel to test the possible effect of electrical current alone. The multibarrel electrode was advanced by a manual micropositioner through the dura and cerebellum to the AVCN. Each penetration was constrained within a small range of stereotaxic angles that resulted in penetration of the AVCN (this range of angles was verified in pilot studies to result in recording sites limited to the AVCN), as determined in Gai and Carney (2006). After recording from a neuron, the electrode was moved by 150 m and a delay of 30 min was imposed before recording from another neuron. Data analysis Neuron response types were categorized based on Blackburn and Sachs (1989): 1) primary-likes (PLs), 2) primary-like-with-notches (PLNs), 3) choppers (including different chopping types), 4) onset, and 5) unusual response types. Unusual response types were those that did not satisfy the criteria for any of the other response categories. During injections, the SR was computed based on spikes within a 50- to 100-ms window after the onset of the 25-ms tones at the lowest level used for the rate-level function. (The sound-evoked response was always within 40 ms of tone onset and the lowest tone level was always below threshold.) The SR obtained with this method agreed with the SR computed from the tuning curve program, which was run when each neuron was first encountered. First-spike latency (FSL) is difficult to measure because of the presence of spontaneous activity. Young et al. (1988) set a cursor by hand at the beginning of sound-evoked activity based on visual examination of the poststimulus time histogram (PSTH) and discarded all activity before the cursor. Chase and Young (2007) provided a binless algorithm that was used here; this algorithm determines the starting time of the sound-evoked activity by computing the probability of an instantaneous firing rate that can or cannot be generated by spontaneous activity. Specifically, the algorithm first searches for the time when at least five spikes (combined across all stimulus repetitions) have occurred. Then the probability of generating a number of spikes in a certain time interval by spontaneous firing is calculated for a set of intervals between the present spike and each previous spike. The shortest interval includes the previous five spikes and the longest interval includes all previous spikes. When the minimum probability across different lengths of intervals is below a certain criterion (i.e., 10 6 ), the time is marked as the start of tone-evoked activity; otherwise, the time point moves to the next spike. In this study, the starting point was used to exclude spontaneous spikes that occurred before the sound-evoked response. The mean FSL was then computed based on the remaining spikes. A t-test was used to test the significance of rate changes in response to CF tones after blocking inhibition and t values were compared across neuron types. In this study, the reported t value is the maximum t value across tone levels unless otherwise specified. Analysis of drug effects In the following text, positive effect (Caspary et al. 1994) refers to consistently increased neural responses to CF tones based on the rate-level functions, frequently repeated during iontophoresis. Drug effect (Caspary et al. 1994) refers to an effect on neural responses that was presumably caused by the inhibitory receptor antagonists or agonists, rather than by other unknown factors (e.g., current injection, mechanical effects of the electrode on the neuron, or change in ph). The rate-level function was frequently obtained (approximately every 5 min) to monitor the change of average firing rate and to check the stability of action potentials, which ensured that recordings were from the same neuron over time. Neurons that showed unchanged rate were counted as negativeeffect neurons if one of the two following conditions was met: 1) the injecting and recording tips of the electrode remained close ( 25 m) to each other after the recording session, as judged with a 100 microscope after the recording session; 2) during the same penetration, there were other positive-effect neurons. If neither of these conditions was met, the neuron was not included in the results presented here. 2 In other words, a neuron that showed no change in 2 The 100 microscope became available after the first nine positive-effect neurons (from experiments g340 g354) had been studied. Before the introduction of the microscope, neurons without a positive effect were excluded from this study unless the same penetration yielded other positive-effect neurons.

5 1080 Y. GAI AND L. H. CARNEY rate was counted as a negative-effect neuron only if a given electrode was known to function properly. A neuron was considered to show a positive effect only when it satisfied the following two criteria: 1) average rate significantly increased (t-test, P 0.05) for at least one superthreshold level ( 10 db above threshold) and 2) the rate increase remained significant until the end of the injection and did not disappear immediately on termination of the injection. (The maximum effect commonly occurred some time after termination of the injection current.) Some neurons occasionally showed significant rate increases during the injection, but the increase disappeared before or immediately after the injection terminated; these were considered to be negative-effect neurons. Recovery of the drug effect was indicated when the average discharge rate consistently decreased after termination of the injection by 10% of the maximum rate. The recovery process was highly variable across neurons and across inhibitory receptor antagonists. For neurons that did not show signs of recovery within the holding time (between 3 min and 4 h), the positive effects were still considered drug effects if the preceding two criteria were met. RESULTS Responses to short tones at CF Responses of 89 neurons in 38 gerbils are presented. Table 1 lists neurons that showed positive effects to inhibitory receptor antagonists. Fifty-four of 89 neurons showed positive effects after injection of either glycine (21/59) or GABA (33/73) receptor antagonists. Of these 54 neurons, 12 were tested with both types of antagonists; these neurons showed a positive effect and full recovery to whichever inhibitory receptor antagonist was tested first for that neuron. Two neurons (2/12, one sustained chopper and one unusual response type) showed positive effects for both antagonists. In general, larger rate changes were observed after blocking GABAergic inhibition than after blocking glycinergic inhibition. (The maximum t values for changes in sound-evoked rate averaged across all neurons were 13.8 and 9.9, respectively.) Thirty of 37 negative-effect neurons were tested with both types of inhibitory receptor antagonists and the other 7 negative-effect neurons were tested with only one type of antagonist. As stated earlier, negative-effect neurons sometimes showed significant but temporary rate increases; however, the increased rate either disappeared before or immediately after termination of the injection. There were 29 to 47% of positive-effect neurons that showed recovery for different inhibitory receptor antagonists (Table 1). TABLE 1. Summary of positive effects for all recorded neurons Although the percentage of neurons that showed full or partial recovery did not differ substantially for different antagonists (Table 1), it was commonly observed that recovery after injection of bicuculline was relatively quick and complete as compared with the other two antagonists. Glycine and GABA were injected only during recordings from two and three neurons, respectively. A chopper that showed a positive effect to bicuculline did not change responses in the presence of glycine (a strychnine injection was not available for this neuron). An unusual response type that showed different positive effects to strychnine and bicuculline had a decreased rate after glycine injection (this neuron will be subsequently described in detail). One PL and one chopper showed positive effects to bicuculline but negative effects to GABA. The responses of the last neuron, a PL, did not change after injection of strychnine, bicuculline, or GABA. Figures 1 5 show examples of different response types in response to the injection of inhibitory receptor antagonists. Figure 6 shows all 54 positive-effect neurons. General observations for each response type follow. PRIMARY-LIKES (PLS) AND PRIMARY-LIKE-WITH-NOTCHES (PLNS). Figure 1 shows changes in rate-level functions (left column), PSTHs (middle columns), and interspike intervals (ISIs) versus time (right columns) for two PLs caused by injecting bicuculline and strychnine, respectively. The PSTHs and interval plots were obtained at the highest tone level (70 or 80 db SPL). The first neuron showed a maximum change 61 min after the 15-min injection of bicuculline ended; the positive effect was highly reduced 166 min after the injection ended. For the second neuron, the maximum change occurred 108 min after termination of the 54-min injection of strychnine. The first PL (Fig. 1A) did not show a change in the shape of the PSTH, consistent with most PLs. The onset gradually became earlier and sharper for the second PL (Fig. 1B) asthe average rate increased. Other neurons also showed this early peak during or after injection; this phenomenon will be subsequently discussed in more detail. The coefficient of variation (CV; Blackburn and Sachs 1989) is shown in the right columns (Fig. 1) to quantify the discharge regularity. The CV of the first PL did not change, whereas the CV of the second PL decreased as the onset peak moved earlier. Since both PLs had CFs between 1 and 2 khz, they phaselocked to the tone frequency. No change in the synchronization coefficient was observed with drug injection. In general, block- Cell Type PL PLN Chp-S Chp-T Chp-SA Unusual Onset Overall Strychnine n Positive effect (35%) 10% recovery (33%) Bicuculline n Positive effect (63%) 10% recovery (47%) Gabazine n Positive effect (35%) 10% recovery (31%) PL, primary-like; PLN, primary-like-with-notch; Chp-S, sustained chopper; Chp-T, transient chopper; Chp-SA, slowly adapting chopper.

6 INHIBITORY EFFECTS ON AUDITORY RESPONSES 1081 ing inhibition did not vary the synchronization to tone frequency or the phase of the synchronized response for PLs. Diverse effects of inhibition were observed when comparing the rate-level functions and PSTHs for all 18 PLs and 3 PLNs (Fig. 6, the first 21 groups of panels). For example, the shape of the rate-level function and the first-spike latency changed in different ways across neurons. CHOPPERS. Figure 2 shows changes in tone responses after gabazine injection for a transient chopper (Chp-T). In addition to an increased average discharge rate, a systematic change in the PSTH was observed. Before the injection of gabazine, three clear chopping cycles could be identified (Fig. 2, top row, middle column). During the injection, a distinct fourth peak was observed (Fig. 2, middle row, middle column). This neuron showed full recovery 18 min postinjection (Fig. 2, bottom row, middle column); the average rate decreased and the fourth peak disappeared. The right column shows the mean ISIs over time that were used to classify chopper types. The intervals were computed over 20 ms to avoid end effects (Blackburn and Sachs 1989; Young et al. 1988). This neuron was classified as a transient chopper [also called a transiently adapting chopper FIG. 1. A: pure-tone responses of a primary-like (PL) response type (g396u1) before and after injection of bicuculline. Total duration of bicuculline injection was 15 min. Left column: rate-level functions for characteristic-frequency (CF) tones (CF 1,808 Hz). Middle and right columns: poststimulus time histograms (PSTHs) and mean interspike intervals (ISIs) as a function of time in response to a 70 db SPL tone at CF. Top row: responses before injection. Middle and bottom rows: responses after injection. CV, coefficient of variation (Blackburn and Sachs 1989). B: PL, g342u4; total duration of strychnine injection was 54 min. CF 1,103 Hz (same format as that in A). Note the appearance of early peaks in the middle panels of the bottom 2 rows. (Blackburn and Sachs 1989)] because the intervals increased abruptly over the early part of the response and then slightly decreased later (Fig. 2, top row, right column). During the injection of gabazine, the abrupt increase of intervals at the beginning of the response disappeared (Fig. 2, middle row, right column). This neuron did not show a positive effect to strychnine injection. Figure 3 shows changes in response to GABA-receptor antagonist injection for two slowly adapting choppers (Chp- SAs). These examples, one for gabazine and the other for bicuculline, are presented to show that some basic observations obtained with these two types of GABA-receptor antagonists were generally consistent, except the recovery process was usually faster after a bicuculline injection. These two neurons were classified as Chp-SAs since their ISIs increased with time. However, when GABA-receptor antagonists were injected, the ISIs became relatively constant with time so that the two slowly adapting choppers indeed became sustained choppers. The CV decreased from 0.26 to 0.09, and from 0.25 to 0.15 for the two choppers, respectively. The change of regularity can also be observed in the PSTHs (middle columns)

7 1082 Y. GAI AND L. H. CARNEY FIG. 2. Transient chopper (Chp-T), g387u12; total duration of gabazine injection was 11 min. CF 2,362 Hz (same format as that in Fig. 1). Note the extra cycles in the middle column, 2nd row. because a larger number of distinct chopping cycles appeared during drug injection. The chopping cycles also decreased correspondingly. The second Chp-SA was also tested with strychnine injection but did not show a positive effect. Figure 4 shows changes after bicuculline injection for a sustained chopper (Chp-S). The ISIs of sustained choppers are nearly constant over time (Blackburn and Sachs 1991). In the present study, various changes to the sustained properties were observed. In the ISI plots (Fig. 4, top row, right column), before blocking inhibition, intervals slightly increased as a function of time. Nevertheless, this neuron was classified as a Chp-S because the increase was substantially smaller than that FIG. 3. A: slowly adapting chopper (Chp- SA), g381u2; total duration of gabazine injection was 12 min. CF 1,547 Hz. B: Chp-SA, g362u2; total duration of bicuculline injection was 23 min. CF 543 Hz (same format as that in Fig. 1). Note the faster chopping and extra cycles in A, 2nd row, middle and in B, 2nd row, middle.

8 INHIBITORY EFFECTS ON AUDITORY RESPONSES 1083 FIG. 4. Sustained chopper (Chp-S), g362u1; total duration of bicuculline injection was 16 min. CF 1,476 Hz (same format as that in Fig. 1). Note the faster chopping in the 2nd row, middle panel. for Chp-SA response types (Chp-S responses had CVs 0.2 and Chp-SA responses had CVs 0.2). After bicuculline was injected, the intervals became more invariant over time. Similar to the Chp-SA response, more chopping cycles with a faster chopping rate were observed in the PSTH (Fig. 4, middle column) of the Chp-S response. This neuron did not show a positive effect to strychnine injection. As shown in the population plot (Fig. 6), the change in discharge rate in response to on-cf tones after the injection of inhibitory receptor antagonists was smaller for Chp-S response types (Fig. 6, #22 29) than for Chp-SA (Fig. 6, #36 40) and Chp-T (Fig. 6, #30 35) response types. Previous studies also suggested that Chp-S response types may not receive strong inhibition (Banks and Sachs 1991; Hewitt and Meddis 1993) or that Chp-S response types mainly receive off-cf inhibition (Paolini et al. 2005). However, since Chp-S response types had more regular discharges, the variance in rate was smaller (confirmed by this study), and thus the same amount of rate change can be more meaningful compared with the other two chopper types. Therefore the t value was used to quantify the significance of changes in discharge rate. In fact, the maximum t values of Chp-S responses across tone levels were comparable to the values of other response types (not shown); it was thus concluded that Chp-S responses to CF tones were also affected by inhibitory inputs. As shown by the two Chp-SA types and one Chp-S type, the regularity of choppers could be affected by inhibitory inputs. Figure 7 shows the CV values before and after the injection of inhibitory receptor antagonists for all choppers that showed positive effects (small CV values indicate regular discharges). The CV is traditionally computed based on discharges in the time window of 12 to 20 ms (Blackburn and Sachs 1989), as shown in Fig. 7 (left). Chp-SA response types (Fig. 7, left, plus symbols) showed the greatest change in CV after blocking inhibition (three of five responses actually became sustained choppers). Six of eight Chp-S response types (Fig. 7, left, right triangles) also showed decreased CV. Although one Chp-T became a Chp-S after injection of GABAergic antagonists, no systematic change in regularity was observed for the other five Chp-T response types (Fig. 7, left, downward triangles). However, as illustrated by the Chp-T in Fig. 2, the change in ISIs primarily occurred before 12 ms, which was excluded by the traditional 12- to 20-ms CV time window. Figure 7 (right) shows the CV based on discharges that occurred between 0 and 12 ms after stimulus onset. All Chp-T response types showed decreased regularity within this time window (Fig. 7, right, downward triangles). In summary, irregular chopping patterns can be generated from regular chopping patterns by inhibitory inputs; however, inhibitory inputs do not necessarily account for all irregular chopping patterns. (This finding is revisited in the DISCUSSION.) UNUSUAL AND ONSET RESPONSE TYPES. Two onset and 12 unusual response types are presented here. A higher percentage of these neurons showed positive effects compared with the two major response types (Table 1). Four unusual response types showed chopper-like or multimodal patterns, but they did not satisfy the one-spike-per-peak criterion required for categorization as choppers (Blackburn and Sachs 1989). Figure 5 shows responses of an unusual response type to bicuculline (second row), strychnine ( fourth row), and glycine (bottom row), respectively. In response to CF tones, this neuron had a relatively broad peak at the response onset, followed by low activity that built up in time (top row, middle column). After injecting bicuculline, both onset and sustained activity increased (second row, middle column) and ISIs decreased (second row, right column). The activity increase was maximal during the latter portion of the response and minimal right after onset. This neuron recovered by 64 min postinjection (third row). Similar to the Chp-T response type illustrated in Fig. 2, the recovered average rate was slightly lower than the control rate at high tone levels (Fig. 5, third row, left column). This overrecovery was reflected in the PSTHs as a less distinct onset and lower activity at the end of the response. The most interesting change in response to strychnine injection was the appearance of a precisely timed early peak (Fig. 5, fourth row, middle column). This peak was 4.7 ms earlier than the broad peak seen in the original response (top row, middle column). There was also a 1- to 1.5-ms-wide notch after the early peak. The sustained activity showed less increase and was constant over time. The average rate recovered 31 min postinjection ( fifth row, left column). The early peak decreased

9 1084 Y. GAI AND L. H. CARNEY FIG. 5. Unusual response type, g368u2; total duration of drug injection was 23, 60, and 7 min for bicuculline, strychnine, and glycine, respectively. CF 3,544 Hz (same format as that in Fig. 1). Note the early peaks in the middle column, 2nd, 4th, and 5th rows. by half, but did not totally disappear ( fifth row, middle column). The last row shows the response of this neuron after injection with glycine. Both the early peak and sustained activity decreased drastically, confirming the presence of glycinergic inhibition. The effects of inhibition on the 11 unusual and 2 onset response types were diverse (Fig. 6, bottom three rows; #42 and 47 were the same neuron shown earlier). Besides the unusual response type described earlier, another neuron (Fig. 6, #50) showed highly isolated early peaks at times when no discharges were present in the control responses. One example (Fig. 6, #48) showed an enhanced, but not earlier, onset compared with the control responses. Another example (Fig. 6, #52) had multiple peaks, similar to a chopper (this cell was not characterized as a chopper because it failed the one-spike-perpeak criterion). After blocking inhibition, the timing of these peaks did not change, but an early peak appeared, and the interval between this extra peak and the second peak was approximately the same as the later intervals. On the other hand, two neurons (Fig. 6, #41 and 51) had a distinct extra chopping peak after the primary chopping peaks, but other parts of the response were essentially unchanged. FIRST-SPIKE LATENCY (FSL) AND EARLY INHIBITION. Figure 8 shows the mean (A and B) and SD (C and D) of the FSL for all types of responses before and after blocking inhibition. (Glycinergic and GABAergic inhibitions are not differentiated in the plots because no difference was observed between the two types of inhibition.) The FSL was computed using responses to the highest tone level tested (60 80 db SPL). The mean FSL decreased significantly for 31 neurons (P 0.05), increased significantly for 8 neurons (P 0.05), and remained unaffected for 15 neurons (Fig. 8A). Figure 8B amplifies the shaded area of Fig. 8A to provide a better view of short FSLs. These two plots show that the mean FSLs of choppers (circles) did not vary substantially after blocking inhibition. The filled symbols in these two plots represent 13 cells that had FSLs that decreased by 0.8 ms after blocking inhibition (14 symbols are filled in the plots of Fig. 8 because one cell showed FSL decrease of 0.8 ms for both inhibitory receptor antagonists);

10 INHIBITORY EFFECTS ON AUDITORY RESPONSES 1085 FIG. 6. Rate-level functions (odd-numbered columns) and PSTHs (even-numbered columns) in response to CF tones for all neurons before and after blocking glycinergic (Gly) or GABAergic (Gab) inhibition. In the rate-level plots, thick lines are average rates before injection and thin lines are maximum average rates during or after injection. In the PSTH plots, top panels are PSTHs before injection and bottom panels are PSTHs for responses with maximum average rates during or after injection. All PSTHs are responses to the highest-level CF tones studied (60 80 db SPL). all of them were PL, unusual, or onset response types. These units accounted for 44% of all PLs, unusual, and onset types. Figure 8, C and D shows SDs of the FSL. All but two of the neurons that had FSLs that decreased by 0.8 ms (filled symbols) had also decreased SDs (Fig. 8C). The FSL SD for all choppers and PLNs was 1 ms (Fig. 8C, bottom left corner, and Fig. 8D); the FSL SD for all PLs was 1 ms (Fig. 8C). The inhibition associated with cells that showed FSL decreases of 0.8 ms after injecting inhibitory receptor antagonists was called early inhibition. [Note that the 0.8-ms criterion was arbitrarily chosen since no clear boundary existed between J Neurophysiol VOL early inhibition and later inhibition (Fig. 8, A and B).] When applying a t-test to the changes in FSL on blocking inhibition, all FSL decreases were significant for cells with early inhibition. As shown in the mean FSL figure (Fig. 8, A and B), there were different degrees of early inhibition. The mean FSLs of some neurons were shortened by about 0.8 ms, whereas much larger time changes were observed for others. It was unclear whether a single underlying mechanism accounted for the large range of FSL shifts. Four of 13 cells that showed early inhibition required 30 min postinjection time to show the maximum effect and only one showed recovery during the 99 MARCH

11 1086 Y. GAI AND L. H. CARNEY FIG. 7. Coefficients of variation (CVs) computed based on activity in different time windows (left, ms; right, 0 12 ms) for choppers before (x-axis) and after (y-axis) blocking inhibition. Responses to both glycinergic and GABAergic receptor antagonists are included. holding time. It should also be noted that blocking glycinergic and GABAergic inhibitions both caused large changes in FSLs, although the sources of these two types of inhibitions are presumably different. Potential mechanisms for the influence of early inhibition will be subsequently discussed. SOUND-EVOKED RATE AND SPONTANEOUS RATE. As shown in Table 1, some neurons within each response type showed a positive effect, whereas others did not. Because a positive effect was always associated with an increased discharge rate, the possibility that positive effects were associated with relatively low baseline rates (due to the presence of inhibition) compared with negative effects was examined (not shown). A large range of discharge rates (from 50 to 600 spikes/s) was observed for both positive- and negative-effect neurons. The five neurons (all of which were choppers) that had the highest rates ( 450 spikes/s) did not show positive effects to either glycine- or GABA-receptor antagonists. However, for neurons that had average rates lower than these choppers, positive effects were observed for both high- and low-rate responses. It can thus be concluded that variations in excitatory inputs determined the range of discharge rates across neurons and inhibitory inputs modulated the rate only within a limited range. Of the 54 positive-effect neurons (i.e., those for which the increase of sound-evoked rate was significant), 7 showed decreased SR, 16 showed unchanged SR (12 of the 16 had SR 0), and 31 showed increased SR. Figure 9A shows the mean values of SR before and after blocking inhibition for both positive-effect (filled symbols) and negative-effect (open symbols) neurons. One interesting observation was that neurons with SR 45 spikes/s before blocking inhibition, which were all PLs or PLNs, never showed positive effects. Ebert and Ostwald (1995a) computed the percentage of decrease of SR and sound-evoked rate by injecting GABA in the VCN. They found that the SR was significantly more reduced than the sound-evoked rate. In the present study, of the FIG. 8. Mean (top row) and SD (bottom row) of first-spike latency (FSL) before and after blocking glycinergic or GABAergic inhibition. A and C show all cells; B and D show the cells within the left-bottom shaded areas of A and C, respectively. Filled symbols are cells for which mean FSL decreased 0.8 ms after blocking inhibition.

12 INHIBITORY EFFECTS ON AUDITORY RESPONSES 1087 FIG. 9. A: spontaneous rate (SR) before and after blocking glycinergic or GABAergic inhibition. B: percentage of increase of SR vs. percentage of increase of the maximum sound-evoked rate across tone levels after blocking inhibition. Only neurons with control SR 5 spikes/s are plotted. C and D: change in spontaneous rate compared with change in sound-evoked rate after blocking inhibition in terms of t values. D is the shaded area in C. Open/filled symbols indicate neurons with negative/positive effects. 31 neurons that showed increased SR after blocking inhibition, 27 showed a larger increase of SR than of sound-evoked activity. Figure 9B shows the percentage of SR increase versus the percentage of maximum sound-evoked-rate increase. Only cells with control SR 5 spikes/s are plotted, for consistency with Ebert and Ostwald (1995a). Based on this result, which does not take the variance of rate into account, it seemed that inhibitory inputs had a stronger effect on spontaneous activity than on tone-evoked activity. Statistical tests of changes in SR incorporate the variability in rate. Figure 9, C and D shows t values based on changes in SR versus t values based on changes in tone-evoked rate after blocking glycinergic or GABAergic inhibition for both positive-effect (filled symbols) and negative-effect (open symbols) neurons (D amplifies the shaded area in C). Because the variance of SR was generally larger than the variance of sound-evoked rate, the t values for the change in SR were smaller than the t values for the change in sound-evoked rate for 91% of neurons with positive effects. Therefore based on this result, inhibitory inputs had a more significant effect on tone-evoked activity than on spontaneous activity. Although the increase of SR for 33% of positive-effect neurons was significant (t 1.96, P 0.05), one must be cautious when evaluating the drug effect on spontaneous activity for three reasons. First, the SRs of some negative-effect neurons also changed (27% showed significantly increased SR and 13% showed significantly decreased SR). Second, it was observed in at least three positive-effect neurons that SR already varied substantially before the injection of inhibitory receptor antagonists. Third, the recovery of SR was not correlated with the recovery of sound-evoked rate (not shown). FREQUENCY RESPONSE AREA. The response area was used to evaluate whether the frequency tuning of inhibitory inputs was aligned with that of excitatory inputs. Caspary et al. (1994) found that there are more cells with on-cf inhibition than off-cf inhibition for both primary-likes and choppers. In general, the present study also observed more on-cf inhibition (including broad inhibition) than off-cf inhibition (25 on-cf or broad inhibition cells compared with 16 off-cf inhibition cells). A unique finding here was that most of the off-cf inhibition was exhibited by PLs or PLNs (n 9), as shown by the example in Fig. 10A. The majority of choppers (12/14) FIG. 10. Frequency response areas for a PL (A) and a Chp-S (B). The tone level was 80 and 70 db SPL for the two neurons, respectively. Shaded areas indicate rate increase after blocking GABAergic inhibition with gabazine. The vertical dotted lines indicate neuron CFs.

13 1088 Y. GAI AND L. H. CARNEY showed on-cf (especially broad) inhibition, as shown by the example in Fig. 10B. The hypothesis in Paolini (2005) that sustained choppers receive lateral inhibition and transient choppers receive on-cf inhibition was not supported by the present study. Four and 2 of 6 sustained choppers received on-cf and off-cf inhibition, respectively; 4 of 4 transient choppers received on-cf inhibition. Responses to complex sounds RESPONSES TO SINUSOIDALLY AMPLITUDE MODULATED (SAM) TONES. Forty-eight positive-effect (based on pure-tone responses) neurons were further tested with SAM-tone stimuli (19 and 30 neurons for glycine- and GABA-receptor antagonists, respectively). Figure 11 shows changes in the rate (r- MTF) and sync (s-mtf) modulation-transfer functions after blocking inhibition for different response types. PLs (n 17) and PLNs (n 2) generally showed flat r-mtfs and relatively flat s-mtfs. Choppers showed low-pass (n 9), high-pass (n 6), band-pass (n 2), or flat (n 1) r-mtfs. Of course, these shapes can depend on the range of modulation frequency tested with respect to the neuron s individual characteristics (and the choice of the range of modulation frequency was limited by the frequency boundaries in the calibration table). Half of the choppers showed band-pass s-mtfs and the other half showed low-pass s-mtfs, with maximum synchronies higher than those of primary-likes. The r-mtfs and s-mtfs of unusual and onset response types had more diverse shapes than the basic shapes described earlier (Fig. 11). After blocking glycinergic or GABAergic inhibition, 16 of 19 and 28 of 30 neurons showed significantly increased rate (P 0.05; Fig. 11, shaded areas in r-mtfs). Different degrees FIG. 11. Rate (odd-numbered columns) and synchrony (even-numbered columns) modulation transfer functions (MTFs) for all neurons before and after blocking glycinergic (Gly) or GABAergic (Gab) inhibition. For rate MTFs, shaded areas indicate increased average rate after blocking inhibition. For sync MTFs, shaded areas indicate decreased synchronization to modulation frequency after blocking inhibition. The y-axis of the rate plots was from 0 to a value slightly greater than the maximum rate.